View Code? Open in Web Editor NEW

PyTorch implementation of Proximal Policy Optimization (PPO) for use with https://github.com/mweiss17/SEVN

License: MIT License

Python 78.70% Jupyter Notebook 21.30%

sevn-model's Introduction

SEVN Model

In this repository you'll find the code used to train the multi-modal agents on SEVN. These agents can take in images, scene-text, and gps to navigate to goal addresses. This repository was forked from this PPO repository.

Installation

To get install all the software for the SEVN-model, simply run this script:

# Create a conda environment for the dependencies (if the sevn environment already exists, don't recreate it)
conda create -n sevn

# Install the SEVN repository and dependencies 
git clone [email protected]:mweiss17/SEVN.git
cd SEVN
pip install -e .
cd ..

# Install the SEVN-model repository and dependencies
git clone [email protected]:mweiss17/SEVN-model.git
cd SEVN-model
pip install -e .
cd ..

# Install the OpenAI Baselines for Atari preprocessing repository and depenencies
git clone https://github.com/openai/baselines.git
cd baselines
pip install -e .
cd ../SEVN-model

Now, let's verify that the install was correct by (briefly) training a model

python main.py --env-name "SEVN-Train-AllObs-Shaped-v1" \
               --custom-gym SEVN_gym \
               --algo ppo \
               --use-gae \
               --lr 5e-4 \
               --clip-param 0.1 \
               --value-loss-coef 0.5 \
               --num-processes 1 \
               --num-steps 256 \
               --num-mini-batch 4 \
               --log-interval 1 \
               --use-linear-lr-decay \
               --entropy-coef 0.01 \
               --comet mweiss17/navi-corl-2019/UcVgpp0wPaprHG4w8MFVMgq7j \
               --seed 0 \
               --num-env-steps 10000000

sevn-model's People